Question Answering on Merged QA HotpotQA, NarrativeQA, WikiHop (test)
63.34AccuracyLATENTGRAPHMEM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LATENTGRAPHMEMBackbone=Qwen3-8B2026.01 | 63.34 | 49.73 | |
| LATENTGRAPHMEMBackbone=SmolLM3-3B2026.01 | 58.64 | 44.16 | |
| LATENTGRAPHMEMBackbone=Qwen2.5-1.5B-Instruct2026.01 | 56.08 | 38.61 | |
| Reasoner-onlyBackbone=Qwen3-8B2026.01 | 54.74 | 52.61 | |
| MemGenBackbone=Qwen3-8B2026.01 | 54.56 | 58.42 | |
| Reasoner-onlyBackbone=Qwen2.5-1.5B-Instruct2026.01 | 52.11 | 34.63 | |
| Reasoner-onlyBackbone=SmolLM3-3B2026.01 | 50.7 | 37.83 | |
| MemGenBackbone=SmolLM3-3B2026.01 | 49.58 | 47.87 | |
| PREMemBackbone=SmolLM3-3B2026.01 | 45.81 | 31.63 | |
| MemGenBackbone=Qwen2.5-1.5B-Instruct2026.01 | 44.03 | 48.11 | |
| PREMemBackbone=Qwen3-8B2026.01 | 43.3 | 30.36 | |
| A-MemBackbone=SmolLM3-3B2026.01 | 29.44 | 48.54 | |
| Mem0Backbone=Qwen2.5-1.5B-Instruct2026.01 | 25.52 | 19.82 | |
| A-MemBackbone=Qwen2.5-1.5B-Instruct2026.01 | 24.78 | 42.37 | |
| THEANINEBackbone=SmolLM3-3B2026.01 | 24.47 | 13.71 | |
| PREMemBackbone=Qwen2.5-1.5B-Instruct2026.01 | 23.4 | 23.73 | |
| THEANINEBackbone=Qwen3-8B2026.01 | 20.91 | 10.54 | |
| A-MemBackbone=Qwen3-8B2026.01 | 18.89 | 39.68 | |
| Mem0Backbone=Qwen3-8B2026.01 | 17.74 | 12 | |
| RAGBackbone=Qwen3-8B2026.01 | 16.43 | 43.74 | |
| Mem0Backbone=SmolLM3-3B2026.01 | 15.88 | 11.53 | |
| THEANINEBackbone=Qwen2.5-1.5B-Instruct2026.01 | 10.63 | 6.22 | |
| RAGBackbone=SmolLM3-3B2026.01 | 7.89 | 36.29 | |
| RAGBackbone=Qwen2.5-1.5B-Instruct2026.01 | 6.87 | 21.19 |